Absci
AI drug-creation company using zero-shot generative models to design de novo antibodies, validated by an in-house wet-lab feedback loop.
1. Core Product / Service
Absci is a "data-first generative AI drug creation" company. Its Integrated Drug Creation™ platform couples generative AI models with a synthetic-biology data engine in a continuous loop: AI proposes antibody candidates, a high-throughput wet lab screens them (reported capacity of billions of cells per week), and the results feed back to retrain the models — going from AI-designed candidate to wet-lab-validated candidate in as little as six weeks [1].
Its flagship model, IgDesign1, is described as the first in vitro-validated inverse folding model for antibody design [2]. Absci also advances its own pipeline: ABS-101 (an IBD antibody in IND-enabling studies) and ABS-201 (hair regrowth / androgenetic alopecia) [1].
2. Target Users & Pain Points
- Pharma / biotech partners who want de novo antibodies against hard targets without screening millions-to-billions of natural candidates.
- Its own pipeline — Absci is both a platform vendor and a drug developer, using internal programs to prove the "AI-designed antibody" narrative.
Pain solved: traditional antibody discovery is a high-cost, low-throughput screening exercise; Absci compresses discovery into a computation-first loop with fast wet-lab validation [1].
3. Competitive Landscape
| Player | Approach | Vs. Absci |
|---|---|---|
| nabla-bio | JAM multimodal generative design + wet lab | Both do de novo antibody design; Nabla is partnership-funded, Absci is public |
| chai-discovery | Chai-1 structure / Chai-2 antibody design | Chai is open-source-first with a ~20% hit-rate claim |
| evolutionary-scale | ESM3 sequence/structure/function model | Platform model licensing vs. integrated drug-creation loop |
| isomorphic-labs | AlphaFold-lineage, Alphabet-funded | Big-tech scale; Absci is an independent public biotech |
Absci's differentiation is the tight AI↔wet-lab feedback loop and its own clinical pipeline, rather than selling model access alone.
4. Unique Observations
- The binder speed/accuracy benchmark: this week's research compared Absci's de novo binder generation against an internal "Claude Science" agent pipeline on speed and accuracy — a live test of whether a frontier LLM agent can match a purpose-built antibody-design stack [local]. The "pollution" problem in de novo antibody design (proving generated binders are genuinely novel, not memorized from training data) is a shared open question for both.
- Compute partnership as a signal: Absci's AMD collaboration (deploying Instinct accelerators + ROCm for antibody models) and its data-engine moat suggest the real barrier in AI drug design is proprietary wet-lab data, not raw model scale [2].
5. Financials / Funding
- IPO: July 2021 on Nasdaq (ticker ABSI), priced $16/share, ~$230M gross proceeds, ~$2B debut valuation [3].
- AMD strategic investment: $20M PIPE (Jan 2025) alongside the collaboration [2].
- July 2025 raise: ~$64M gross ($50M underwritten offering + ~$14M ATM) [1].
- Cash: $117.5M as of June 30, 2025; runway guided into H1 2028 [1].
- Revenue: Q2 2025 $0.6M; FY2024 $4.5M; FY2024 net loss $103.1M [1].
6. People & Relationships
- Founder / CEO: Sean McClain (founded 2011 in a Portland basement lab, later Vancouver, WA) [3].
- Partners: AMD (compute + investment), Merck, Astellas [3].
- Peers: nabla-bio, chai-discovery, evolutionary-scale, isomorphic-labs, recursion-pharmaceuticals.
Sources
- [1] Nasdaq, "Absci Reports Business Updates and Second Quarter 2025 Financial and Operating Results" (2026-08-24)
- [2] Absci–AMD collaboration and strategic investment announcement (2026-08-24)
- [3] GeekWire, "Absci shares spike 30% in IPO debut…" (2021) (2026-08-24)